uncomplicate

Björn Ebbinghaus 2017-04-09T12:49:24.826160Z

@blueberry This sounds great!

Björn Ebbinghaus 2017-04-09T14:41:38.067175Z

@blueberry When will 0.10 be released to clojars?

2017-04-09T16:52:37.391985Z

@mroerni No definitive plans, since the updates are still minor. Clone the repository and lein install in the meantime.

Björn Ebbinghaus 2017-04-09T17:13:28.444241Z

@blueberry Okay. By the way, the "fast primitive versions of map and reduce" link in the "Optimized for Clojure" section on http://neanderthal.uncomplicate.org/ ist broken. It points to a localhost url.

Björn Ebbinghaus 2017-04-09T18:07:30.578174Z

A support for primitive arrays would be great...

2017-04-09T18:39:50.658588Z

Primitive arrays open more problems in this area than they solve.

Björn Ebbinghaus 2017-04-09T20:26:20.935245Z

Why? Shouldn't it be easy to copy e.g. an int-array into a ByteBuffer? I have an integer array of 81 million entrys. (dge) doesn't only work with a RealGEMatrix or a clojure seq. If I make a clojure seq out of that array first, my memory will run out. I tried to implement another transit! method, where the source is a double-array. There I want to put the content directly into the RealGEMatrix. I'm facing some problems, because I'm a freshman in clojure or rather java interop.

2017-04-09T20:58:40.018195Z

@mroerni Ah, if by support you mean support copying into native matrices/vectors, then you are right: it is both easy and a good idea. My comment was that support for arrays for the matrix operations themselves is full of landmines and ultimately not able to approach the current speed by a large margin...

Björn Ebbinghaus 2017-04-09T21:00:09.022205Z

Ah, ok. My fault.

2017-04-09T21:27:05.089716Z

@mroerni no fault at all. Please open an issue and describe your use case for primitive array transfer, and I'll support it in the next release.